Parallel Extensions to the.NET Framework Daniel Moth Microsoft

Threading 101 On Single Core Machine –Don’t block the UI Thread Affinity –Async Operations –Synchronization Issues On Multi-core Machine –As above plus Improve Actual Performance

The Manycore Shift

void MultiplyMatrices(int size, double[,] m1, double[,] m2, double[,] result) { for (int i = 0; i < size; i++) { for (int j = 0; j < size; j++) { result[i, j] = 0; for (int k = 0; k < size; k++) { result[i, j] += m1[i, k] * m2[k, j]; } Example Code to Parallelize Matrix multiplication (sequential)

int N = size; int P = 2 * Environment.ProcessorCount; int Chunk = N / P; // size of a work chunk ManualResetEvent signal = new ManualResetEvent(false); int counter = P; // use a counter to reduce kernel transitions for (int c = 0; c < P; c++) { // for each chunk ThreadPool.QueueUserWorkItem(o => { int lc = (int)o; for (int i = lc * Chunk; // process one chunk i < (lc + 1 == P ? N : (lc + 1) * Chunk); // respect upper bound i++) { // original loop body for (int j = 0; j < size; j++) { result[i, j] = 0; for (int k = 0; k < size; k++) { result[i, j] += m1[i, k] * m2[k, j]; } if (Interlocked.Decrement(ref counter) == 0) { // efficient interlocked ops signal.Set(); // and kernel transition only when done } }, c); } signal.WaitOne(); Error Prone High Overhead Tricks Static Work Distribution Knowledge of Synchronization Primitives Heavy Synchronization Lack of Thread Reuse

Parallel Extensions to the.NET Framework.NET library Supports parallelism in any.NET language, without changes to compilers or the CLR Begins to move concurrency from experts to mainstream.NET developers

void MultiplyMatrices(int size, double[,] m1, double[,] m2, double[,] result) { for (int i = 0; i < size; i++) { for (int j = 0; j < size; j++) { result[i, j] = 0; for (int k = 0; k < size; k++) { result[i, j] += m1[i, k] * m2[k, j]; } Matrix multiplication (Seq)

void MultiplyMatrices(int size, double[,] m1, double[,] m2, double[,] result) { Parallel.For(0, size, i => { for (int j = 0; j < size; j++) { result[i, j] = 0; for (int k = 0; k < size; k++) { result[i, j] += m1[i, k] * m2[k, j]; } } }); } Matrix multiplication (Par)

Sample: Raytracer

AGENDA for Remainder Session Task and friends Imperative Task parallelism Parallel class Imperative Data parallelism Parallel LINQ (PLINQ) Declarative Data parallelism

Task class The lightweight unit of work May run concurrently if it’s “worth it” Supports –parent/child relations –explicit grouping –cancelation –waiting

From Threads to Tasks

TaskManager class Represents a scheduler group –think of it as a fancy thread-pool –one implicit to each AppDomain –can create explicitly for isolation –configurable

“Work Stealing” in Action Worker Thread 1 Worker Thread p Program Thread Task 1 Task 2 Task 3 Task 5 Task 4

Imperative Data Parallelism Parallel class –static (overloaded) methods –helper methods to create/work with Tasks –encapsulates common patterns

Parallel.For

Parallel Statements Program statements… …when independent, can be parallelized Synchronous (same as loops) StatementA(); StatementB(); StatementC(); StatementA(); StatementB(); StatementC(); Parallel.Do( () => StatementA(), () => StatementB(), () => StatementC() ); Parallel.Do( () => StatementA(), () => StatementB(), () => StatementC() );

Declarative Data Parallelism Parallel LINQ-to-Objects –Enables LINQ devs to leverage manycore –Fully supports all.NET standard query operators –Minimal impact to existing LINQ model var q = from p in people where p.Name == queryInfo.Name && p.State == queryInfo.State && p.Year >= yearStart && p.Year <= yearEnd orderby p.Year ascending select p; var q = from p in people where p.Name == queryInfo.Name && p.State == queryInfo.State && p.Year >= yearStart && p.Year <= yearEnd orderby p.Year ascending select p;

Parallel LINQ (PLINQ)

PLINQ – Modes of Consumption Pipelined: separate consumer thread –Default for GetEnumerator() And hence foreach loops –More synchronization overhead Stop-and-go: consumer helps –Sorts, ToArray, ToList, GetEnumerator(false), etc. –Minimizes context switches But higher latency and more memory Inverted: no merging needed –ForAll extension method –Most efficient by far But more difficult (side-effects) Thread 2 Thread 4 Thread 1 Thread 3 Thread 1 Thread 3 Thread 1 Thread 2 Thread 1 Thread 3 Thread 2 Thread 1

Summary – Parallel Extensions Task Parallel Library (TPL) –Imperative Task Parallelism – TaskManager etc –Imperative Data Parallelism – Parallel Parallel LINQ (PLINQ) –Declarative Data Parallelism – AsParallel Coordination Data Structures (CDS) –Thread-safe collections, Work exchange constructs, Phased operation primitives, Locks, Initialization support

Resources Binaries, Code, Forums, blogs, videos, screencasts, podcasts, articles, samples My blog for resources to this session

© 2008 Microsoft Corporation. All rights reserved. Microsoft, Windows, Windows Vista and other product names are or may be registered trademarks and/or trademarks in the U.S. and/or other countries. The information herein is for informational purposes only and represents the current view of Microsoft Corporation as of the date of this presentation. Because Microsoft must respond to changing market conditions, it should not be interpreted to be a commitment on the part of Microsoft, and Microsoft cannot guarantee the accuracy of any information provided after the date of this presentation. MICROSOFT MAKES NO WARRANTIES, EXPRESS, IMPLIED OR STATUTORY, AS TO THE INFORMATION IN THIS PRESENTATION.